►
Description
Walkthrough of the UI integration overview
https://gitlab.com/groups/gitlab-org/modelops/applied-ml/review-recommender/-/epics/3
A
Hello,
everybody,
my
name,
is
tan
from
apply
ml
team.
Today
I
would
like
to
work
through
the
ui
integrations
of
suggested
reviewer
with
gitlab,
so
the
presentations
will
basically
split
into
two
parts.
The
first
part
will
be
kind
of
a
high
level
overview.
A
So
let's
get
start
for
the
suggested.
Reviewer
integrations.
There
are
three
major
flow
one.
The
first
one
is
onboarding.
Second
suggestions
and
the
third
one
is
a
feedback
flow
okay.
So
maybe
a
quick
recap
on
the
current
experience
so
currently
for
onboarding
well
well,
actually,
at
the
moment
we
do
onboard
better
customer.
A
We
we,
we
basically
have
to
do
a
bunch
of
manual
steps
to
get
them
on
board,
so
the
first
is
to
to
be
done
by
the
dev
team.
So
we
identify
the
project
and
namespace
of
the
beta
customer,
and
then
we
create
a
pipeline
schedule
that
points
to
the
project
namespace.
A
This
is
integrated
with
the
ci
template
right,
so
it's
executed
as
part
the
ci.
So
when
the
pull
request
is
created,
they
will
be
sent
through
to
to
our
suggest
reviewer
system
and
at
the
end
of
it
we
generate
an
artifacts
and
get
embedded
as
a
comment
on
the
mr
page.
A
A
Okay
to
move
into
what
we
looked
to
do
next
right.
So
for
the
onboarding
we
keen
to
get
you
know
the
user
to
self-serve,
so
some
way
of
ability
to
enable
suggest
your
view
of
your
settings
and-
and
you
know
when
they
click
on
the
buttons
on
the
project
setting
page,
for
example,
they
will
just
kick
off
a
bunch
of
downstream
system
to
get
the
model
built
and
permission
set,
and
all
that
that's
for
the
suggestions
flow.
A
We
we
want
to
focus
on
the
integrated
with
the
ui
so
definitely
get
through
to
the
mri
sidebar.
A
We
hook
into
the
api
calls
to
suggest,
with
your
back
end,
for
the
feedback
flow,
we're
going
to
look
into
collect
some
telemetry
around
user
interaction
with
suggested
reviewer
and
use
that
information
to
incorporate
the
future
model
building,
and
things
like
that.
So
so.
A
So
here
is
kind
of
a
bro
laid
out
what
we
think
you
know
what
we
try
to
do
in
terms
of
the
ui,
but
I
mean
this
is
subjected
to
change
and
this
is
really
really
a
high
level.
I
welcome
you
to
have
a
look
at
the
design
issues
that
I
can
kind
of
walk
you
through
a
little
bit
later
to
get
more
ideas,
but
it
is
still
refining
and
and
so
on.
A
So
don't
kind
of
stick
to
this,
as
at
the
moment
is
really
kind
of
just
give
you
an
idea
of
how
things
are
kind
of
kind
of
lay
out,
but
a
lot
of
is
going
to
be
changing,
okay,
so
the
the
ui.
Basically
we're
going
to
be.
You
know,
project
settings
somewhere
in
the
project
setting
page
we'll
get
a
suggested
view
of
sections
with
the
buttons
to
do
various,
depending
on
the
states
of
the
customer
on
how
they're
on
boarding
with
suggested
viewer.
A
We
show
different
ui,
so
at
the
moment,
if
they
haven't
actually
onboard
yet
potentially
inactivate
buttons
with
some
informations
about
you
know
what
involved,
what
data
we
collect
and
so
on
so
forth.
So
so
that's,
but
if
they're
on
boarding
ready,
they
might
just
have
to
show
a
different
state
on
the
ml
pages
after
the
onboarding.
We're
gonna
kind
of
you
know.
Do
the
api
call
and
make
sure
that
on
the
email
page
they
see
some
suggestions
coming
up
when
they
view
the
mr.
A
Page
or
they
could
be
in
the
creations
or
it
could
be
an
edit
page
or
both,
so
so
that
and
for
the
feedback
we
will
try
to
collect
a
number
of
events
that
related
to
their
journey,
with
the
gesture
viewer,
to
try
to
build
a
phone
or
of
you
know
the
usage
and
understand
how
they
or
how
also
some
of
that
will
reflect
how
effective
the
model
is,
for
example,
if
we
present
it
as
suggestions
and
how
many
of
them
get
accepted
and
how
many
of
them
get
rejected
for
some
reason.
A
So
so
that's
also
coming
handy
as
well.
We
do
analysis
for
the
model
iterations
all
right,
so
at
high
level,
you
can
see
that
it's
all
connected
to
git
lab
it's
all
part
of
the
gitlab
integration.
So
let's
go
go
and
stare.
We
have
a
client
get
lab
that
call
out
to
the
bot
service,
so
this
is
kind
of
a
a
gateway
to
proxy
into
our
kind
of
internal
model,
serving
architecture
and
all
that,
so
the
bot
servers
will
make
yeah.
A
We
have
their
own
data
stored
and
we'll
talk
to
the
model
serving
to
retrieve
the
model
and
respond
to
various
different
interactions
from
get
lab
point
of
view.
A
So
the
next
sessions
are
gonna
kind
of
zoom
out
this
and
then
just
show
you
a
little
bit
more
into
what
involved
in
this
interaction
between
different
systems
and
hopefully
will
give
you
a
better
idea.
Okay,.